Skip to main content
Add a name or technical term to Vocabulary when Utter repeatedly spells it incorrectly. Use text replacements when you need a specific recognized phrase changed to another phrase. Vocabulary supplies useful words to transcription; it does not guarantee a spelling or train a personalized speech model. Results depend on the selected voice model and recording quality.

Add a Custom Word

1

Open Vocabulary

Open the main Utter window and choose Vocabulary in the sidebar.
2

Enter the Preferred Spelling

Under Vocabulary Words, type the word in Enter word, name, or term. For example, enter PostgreSQL rather than a phonetic spelling.
3

Add the Word

Press Enter or click the plus button. Check that the word appears in the list and is active.

Test the Result

Make a new recording with the same mode that produced the mistake. Say a natural sentence such as “Send the PostgreSQL migration notes to Priya.” Check the original transcription and any AI-rewritten output separately. What to expect: The saved word is available to subsequent transcription workflows. Adding vocabulary does not rewrite old transcripts. A rewritten result may differ from the original because the mode also applies AI instructions.

Choose Vocabulary or Text Replacement

  • Vocabulary Words: Add the spelling you want recognized, such as a colleague’s name, product name, or project acronym.
  • Text replacements/snippets: Use a source phrase and its replacement when a recurring recognized phrase should become specific output. Test in the mode where you plan to use it.
  • Mode instructions: Use AI modes for broader formatting or tone changes, not as a substitute for checking names.
Keep vocabulary entries focused. Do not append “sounds like…” instructions to a word: enter only the spelling or phrase you want in the result. Avoid adding several contradictory spellings for the same name.

If the Word Is Still Wrong

  1. Confirm the entry is active and test a new recording.
  2. Check the selected voice language and voice model.
  3. Compare original and enhanced output to find whether the error occurs during transcription or rewriting.
  4. Try the name in a complete sentence with a clear microphone signal. See audio-quality troubleshooting.

Still Need Help?

Contact support with the intended spelling, the result you received, your selected model, and Utter version. Use a non-sensitive example; do not share confidential names or recordings.